Plant disease detection and classification techniques: A review

Abstract

Crop production can be greatly affected by different diseases, which seriously threaten food security. Consequently, detecting plant diseases earlier and preventing the spread of plant diseases is necessary to avoid the economic imbalance. Nevertheless, the manual detection of plant diseases is a time-consuming and error-prone process. Numerous existing techniques are adopted that are unreliable in terms of accuracy and fail to identify the infected region due to non-uniform complex backgrounds resulting in mispredictions. Hence, this research presents a review that focuses on enhancing plant disease detection accuracy and early intervention. The comparison of various machine learning and deep learning techniques, data acquisition, segmentation methods, and feature extraction techniques are presented, particularly in disease prediction. The basic prediction approach encloses the classification model which is trained and the publicly available dataset or the collection of real-time data from the farms is tested. Those datasets include images of normal as well as plants with disease spots which serve as benchmarking datasets for the research. The paper recognizes challenges like limited data, scalability, and accuracy but seeks to leverage previous technologies to advance agricultural practices. Ultimately, the aim is to improve upon existing methodologies for more effective plant disease prediction, contributing to a more robust agricultural sector.